@tea-agent/loop-agent 0.35.1-beta.0 → 0.35.1-beta.2

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (179) hide show
  1. package/AGENTS.md +110 -108
  2. package/CHANGELOG.md +24 -26
  3. package/README.md +165 -165
  4. package/bin/agent-worker.js +0 -0
  5. package/bin/loop-agent.js +57 -21
  6. package/dist/application/task-lifecycle/advance.js +0 -1
  7. package/dist/build-stamp.json +6 -0
  8. package/dist/cli/program.js +2 -2
  9. package/dist/commands/cursor-prompt.js +6 -6
  10. package/dist/commands/init-upgrade.js +19 -351
  11. package/dist/commands/init.js +67 -14
  12. package/dist/commands/loop-benchmark.js +11 -11
  13. package/dist/commands/pi-reuse-benchmark.js +16 -16
  14. package/dist/commands/run-dag-progress.js +0 -14
  15. package/dist/commands/task-advance.js +3 -33
  16. package/dist/executors/dag-pi-executor.js +44 -0
  17. package/dist/shared/operator/capabilities.js +1 -38
  18. package/dist/shared/package-metadata.js +42 -0
  19. package/dist/sidecars/cursor-prompt/executor.js +1 -1
  20. package/dist/worker/console/chat/pi-runtime.js +25 -41
  21. package/dist/worker/console/chat/routes.js +4 -27
  22. package/dist/worker/console/operation-runner.js +0 -24
  23. package/dist/worker/console/operator-actions.js +0 -58
  24. package/dist/worker/console/static/assets/index-CvsQgALl.js +56 -0
  25. package/dist/worker/console/static/assets/{index-Dups4sSM.css → index-hJqCPs_g.css} +1 -1
  26. package/dist/worker/console/static/index.html +2 -2
  27. package/dist/worker/console/static-src/app/useRecoveryConsole.js +5 -0
  28. package/dist/worker/console/static-src/operator-chat/useChatSessions.js +2 -13
  29. package/dist/worker/console/static-src/operator-chat/useComposer.js +7 -30
  30. package/dist/worker/loop-agent/loop-agent-client.js +17 -3
  31. package/dist/worker/observability/read-model.js +20 -0
  32. package/dist/worker/observe/static/copy.js +67 -67
  33. package/dist/worker/observe/static/dag-layout.d.ts +36 -36
  34. package/dist/worker/observe/static/dom.js +220 -220
  35. package/dist/worker/observe/static/relations.js +133 -133
  36. package/dist/worker/observe/static/run-processing.js +148 -148
  37. package/dist/worker/observe/static/views/batch.js +227 -227
  38. package/dist/worker/observe/static/views/failures.js +143 -143
  39. package/dist/worker/observe/static/views/feature.js +492 -492
  40. package/dist/worker/observe/static/views/run.js +453 -453
  41. package/dist/worker/observe/static/views/shell.js +7 -7
  42. package/dist/worker/observe/static/views/timeline.js +163 -163
  43. package/dist/worker/preflight.js +2 -1
  44. package/dist/workflows/dag/backend-test-scenario-param.js +33 -23
  45. package/dist/workflows/dag/canvas-observer.js +275 -275
  46. package/dist/workflows/dag/contract-output-registry.js +14 -0
  47. package/dist/workflows/dag/contract-validator-registrations.js +8 -0
  48. package/dist/workflows/dag/dynamic-runtime/shared.js +9 -1
  49. package/dist/workflows/dag/frontend-implementation-contract.js +233 -39
  50. package/dist/workflows/dag/frontend-prewrite-gate.js +364 -61
  51. package/dist/workflows/dag/frontend-recovery-plan.js +73 -0
  52. package/dist/workflows/dag/frontend-recovery-root-manifest.js +123 -0
  53. package/dist/workflows/dag/frontend-recovery-run.js +539 -0
  54. package/dist/workflows/dag/frontend-repair.js +219 -18
  55. package/dist/workflows/dag/frontend-verification-trace.js +47 -32
  56. package/dist/workflows/dag/frontend-writer-recovery.js +106 -0
  57. package/dist/workflows/dag/frontend-writer-rollback.js +821 -0
  58. package/dist/workflows/dag/init-hybrid.js +41 -24
  59. package/dist/workflows/dag/node-execution.js +89 -0
  60. package/dist/workflows/dag/recovery-recommendation.js +58 -0
  61. package/dist/workflows/dag/runner.js +245 -11
  62. package/dist/workflows/dag/scheduler.js +257 -3
  63. package/dist/workflows/dag/types.js +130 -2
  64. package/docs/architecture/evolution.md +73 -73
  65. package/docs/architecture/system-overview.md +100 -100
  66. package/docs/architecture/worker-and-feature.md +122 -122
  67. package/docs/skills/README.md +7 -7
  68. package/docs/templates/adr.md +60 -60
  69. package/docs/templates/agent-dag-authority-surface-audit.prompt.md +94 -94
  70. package/docs/templates/agent-dag-decision-envelope.schema.json +213 -213
  71. package/docs/templates/agent-dag-decision-gate.prompt.md +246 -246
  72. package/docs/templates/agent-dag-process-supervisor.prompt.md +98 -98
  73. package/docs/templates/agent-dag-report.schema.json +473 -473
  74. package/docs/templates/agent-dag-review-verdict.prompt.md +68 -68
  75. package/docs/templates/backend-test-result.schema.json +99 -99
  76. package/docs/templates/evaluation/agents-map-slim-v1.md +87 -87
  77. package/docs/templates/evaluation/agents-map-verbose-v0.md +153 -153
  78. package/docs/templates/feature-spec.md +53 -53
  79. package/docs/templates/frontend-design-contract.md +42 -42
  80. package/docs/templates/frontend-eval/fixtures/failures/01-type-build-error.md +17 -17
  81. package/docs/templates/frontend-eval/fixtures/failures/02-unit-component-test-fail.md +16 -16
  82. package/docs/templates/frontend-eval/fixtures/failures/03-fixture-schema-drift.md +16 -16
  83. package/docs/templates/frontend-eval/fixtures/failures/04-missing-loading-empty-error-state.md +16 -16
  84. package/docs/templates/frontend-eval/fixtures/failures/05-forbidden-write-writeset-expansion.md +16 -16
  85. package/docs/templates/frontend-eval/fixtures/failures/06-unapproved-dependency-add.md +16 -16
  86. package/docs/templates/frontend-eval/fixtures/failures/07-mock-production-on.md +21 -21
  87. package/docs/templates/frontend-eval/fixtures/functional/01-simple-component-style.md +29 -29
  88. package/docs/templates/frontend-eval/fixtures/functional/02-form-validation.md +28 -28
  89. package/docs/templates/frontend-eval/fixtures/functional/03-list-detail-page.md +28 -28
  90. package/docs/templates/frontend-eval/fixtures/functional/04-api-mock.md +29 -29
  91. package/docs/templates/frontend-eval/fixtures/functional/05-permission-auth-gated-ui.md +27 -27
  92. package/docs/templates/frontend-eval/fixtures/functional/06-ssr-server-client-boundary.md +28 -28
  93. package/docs/templates/frontend-eval/fixtures/functional/07-shared-public-component-api.md +28 -28
  94. package/docs/templates/frontend-eval/fixtures/functional/08-pure-local-no-remote.md +27 -27
  95. package/docs/templates/frontend-eval/metrics.md +138 -138
  96. package/docs/templates/frontend-eval/smoke-targets.md +53 -53
  97. package/docs/templates/frontend-task-constraints.md +35 -35
  98. package/docs/templates/frontend-task-requirement.md +70 -70
  99. package/docs/templates/init-evolution-review.md +35 -35
  100. package/docs/templates/init-managed-agents.md +154 -156
  101. package/docs/templates/interactive-ui-round2-experiment.md +66 -66
  102. package/docs/templates/knowledge-graph-bootstrap-dag.json +118 -118
  103. package/docs/templates/knowledge-sync-dag.json +178 -178
  104. package/docs/templates/knowledge-sync-draft.schema.json +71 -71
  105. package/docs/templates/product-line/closeout.yaml +9 -9
  106. package/docs/templates/product-line/design.md +13 -13
  107. package/docs/templates/product-line/links.md +10 -10
  108. package/docs/templates/product-line/requirement.md +17 -17
  109. package/docs/templates/product-line/test-plan.md +7 -7
  110. package/docs/templates/project-start-checklist.md +9 -9
  111. package/docs/templates/qa-report.md +48 -48
  112. package/docs/templates/sprint-contract.md +29 -29
  113. package/docs/templates/worker-dogfood-evidence.md +80 -80
  114. package/docs/templates/worker-dogfood-setup.md +68 -68
  115. package/harness.json +2 -5
  116. package/package.json +2 -2
  117. package/scripts/kb-bootstrap-init-skeleton.sh +0 -0
  118. package/scripts/kb-graph-incremental-prepare.mjs +0 -0
  119. package/scripts/kb-graph-materialize.mjs +105 -105
  120. package/scripts/kb-graph-promote.mjs +164 -164
  121. package/scripts/kb-query.mjs +554 -554
  122. package/skills/agent-worker/SKILL.md +48 -48
  123. package/skills/agent-worker/references/agent-worker-operator.md +159 -159
  124. package/skills/ai-engineering-context/SKILL.md +48 -48
  125. package/skills/analyze-product-dependencies/scripts/test-validators.mjs +0 -0
  126. package/skills/analyze-product-dependencies/scripts/validate-api-documentation.mjs +0 -0
  127. package/skills/analyze-product-dependencies/scripts/validate-dependency-analysis.mjs +0 -0
  128. package/skills/analyze-product-dependencies/scripts/validate-product-requirement-input.mjs +0 -0
  129. package/skills/analyze-product-requirements/scripts/compute-source-identity.mjs +0 -0
  130. package/skills/analyze-product-requirements/scripts/test-validators.mjs +0 -0
  131. package/skills/analyze-product-requirements/scripts/validate-product-analysis.mjs +0 -0
  132. package/skills/analyze-product-requirements/scripts/validate-product-requirement.mjs +0 -0
  133. package/skills/analyze-product-requirements/scripts/validate-requirement-clarification.mjs +0 -0
  134. package/skills/browser-tools/browser-content.js +103 -103
  135. package/skills/browser-tools/browser-cookies.js +35 -35
  136. package/skills/browser-tools/browser-eval.js +53 -53
  137. package/skills/browser-tools/browser-hn-scraper.js +108 -108
  138. package/skills/browser-tools/browser-nav.js +44 -44
  139. package/skills/browser-tools/browser-pick.js +162 -162
  140. package/skills/browser-tools/browser-screenshot.js +34 -34
  141. package/skills/browser-tools/browser-start.js +86 -86
  142. package/skills/browser-tools/package-lock.json +2556 -2556
  143. package/skills/browser-tools/package.json +19 -19
  144. package/skills/code-review-core/SKILL.md +20 -20
  145. package/skills/codebase-scout/SKILL.md +19 -19
  146. package/skills/grill-me/SKILL.md +10 -10
  147. package/skills/local-jacoco-coverage/scripts/run-coverage-analysis.sh +0 -0
  148. package/skills/local-jacoco-coverage/scripts/start-jacoco-agent.sh +0 -0
  149. package/skills/loop-agent/SKILL.md +0 -1
  150. package/skills/loop-agent/references/command-reference.md +639 -641
  151. package/skills/loop-agent/references/docs-converge.md +126 -126
  152. package/skills/loop-agent/references/learned/README.md +21 -21
  153. package/skills/loop-agent/references/pi-prompt.md +23 -23
  154. package/skills/loop-agent/references/pi-subagent-assisted-mode.md +84 -84
  155. package/skills/playwright-cli/references/element-attributes.md +23 -23
  156. package/skills/playwright-cli/references/playwright-tests.md +39 -39
  157. package/skills/playwright-cli/references/request-mocking.md +87 -87
  158. package/skills/playwright-cli/references/running-code.md +241 -241
  159. package/skills/playwright-cli/references/session-management.md +225 -225
  160. package/skills/playwright-cli/references/storage-state.md +275 -275
  161. package/skills/playwright-cli/references/test-generation.md +433 -433
  162. package/skills/requesting-code-review/SKILL.md +101 -101
  163. package/skills/requesting-code-review/code-reviewer.md +168 -168
  164. package/skills/systematic-debugging/CREATION-LOG.md +119 -119
  165. package/skills/systematic-debugging/condition-based-waiting-example.ts +158 -158
  166. package/skills/systematic-debugging/condition-based-waiting.md +115 -115
  167. package/skills/systematic-debugging/defense-in-depth.md +122 -122
  168. package/skills/systematic-debugging/find-polluter.sh +63 -63
  169. package/skills/systematic-debugging/root-cause-tracing.md +169 -169
  170. package/skills/systematic-debugging/test-academic.md +14 -14
  171. package/skills/systematic-debugging/test-pressure-1.md +58 -58
  172. package/skills/systematic-debugging/test-pressure-2.md +68 -68
  173. package/skills/systematic-debugging/test-pressure-3.md +69 -69
  174. package/skills/using-git-worktrees/SKILL.md +215 -215
  175. package/skills/verification-before-completion/SKILL.md +154 -154
  176. package/skills/webapp-testing/SKILL.md +19 -19
  177. package/dist/worker/console/operation-wait.js +0 -241
  178. package/dist/worker/console/static/assets/index-SjjjZnV3.js +0 -56
  179. package/dist/worker/console/static-src/operator-chat/slash-palette-nav.js +0 -141
@@ -195,280 +195,280 @@ export function createDagCanvasObserver(opts) {
195
195
  flush: () => writer.flush(),
196
196
  };
197
197
  }
198
- const CANVAS_HEADER = `/* AUTO-GENERATED by . run-dag canvas observer. Derived view only — canonical state lives in .harness/dag-runs/. */
199
- import {
200
- Card,
201
- CardBody,
202
- CardHeader,
203
- Divider,
204
- Grid,
205
- H1,
206
- H2,
207
- Pill,
208
- Row,
209
- Stack,
210
- Stat,
211
- Text,
212
- computeDAGLayout,
213
- useHostTheme,
214
- } from 'cursor/canvas';
215
- import { useMemo } from 'react';
216
-
217
- type TaskStatus = 'PENDING' | 'RUNNING' | 'FINISHED' | 'ERROR' | 'SKIPPED';
218
- type Complexity = 'HIGH' | 'MED' | 'LOW';
219
- type Executor = 'cursor' | 'pi' | 'shell' | 'static';
220
-
221
- interface TaskState {
222
- id: string;
223
- depends_on: string[];
224
- complexity: Complexity;
225
- executor: Executor;
226
- subtask_prompt: string;
227
- status: TaskStatus;
228
- model: string;
229
- startedAt?: number;
230
- finishedAt?: number;
231
- resultText?: string;
232
- errorMessage?: string;
233
- inputTokens?: number;
234
- outputTokens?: number;
235
- durationMs?: number;
236
- skippedReason?: string;
237
- }
238
-
239
- interface RunState {
240
- title: string;
241
- runId: string;
242
- startedAt: number;
243
- finishedAt?: number;
244
- runOutcome?: 'SUCCESS' | 'FAILED' | 'INTERRUPTED' | 'PAUSED';
245
- runMessage?: string;
246
- tasks: TaskState[];
198
+ const CANVAS_HEADER = `/* AUTO-GENERATED by . run-dag canvas observer. Derived view only — canonical state lives in .harness/dag-runs/. */
199
+ import {
200
+ Card,
201
+ CardBody,
202
+ CardHeader,
203
+ Divider,
204
+ Grid,
205
+ H1,
206
+ H2,
207
+ Pill,
208
+ Row,
209
+ Stack,
210
+ Stat,
211
+ Text,
212
+ computeDAGLayout,
213
+ useHostTheme,
214
+ } from 'cursor/canvas';
215
+ import { useMemo } from 'react';
216
+
217
+ type TaskStatus = 'PENDING' | 'RUNNING' | 'FINISHED' | 'ERROR' | 'SKIPPED';
218
+ type Complexity = 'HIGH' | 'MED' | 'LOW';
219
+ type Executor = 'cursor' | 'pi' | 'shell' | 'static';
220
+
221
+ interface TaskState {
222
+ id: string;
223
+ depends_on: string[];
224
+ complexity: Complexity;
225
+ executor: Executor;
226
+ subtask_prompt: string;
227
+ status: TaskStatus;
228
+ model: string;
229
+ startedAt?: number;
230
+ finishedAt?: number;
231
+ resultText?: string;
232
+ errorMessage?: string;
233
+ inputTokens?: number;
234
+ outputTokens?: number;
235
+ durationMs?: number;
236
+ skippedReason?: string;
237
+ }
238
+
239
+ interface RunState {
240
+ title: string;
241
+ runId: string;
242
+ startedAt: number;
243
+ finishedAt?: number;
244
+ runOutcome?: 'SUCCESS' | 'FAILED' | 'INTERRUPTED' | 'PAUSED';
245
+ runMessage?: string;
246
+ tasks: TaskState[];
247
247
  }`;
248
- const CANVAS_BODY = String.raw `const NODE_W = 200;
249
- const NODE_H = 72;
250
-
251
- function pillToneFor(status: TaskStatus): 'neutral' | 'info' | 'success' | 'warning' {
252
- switch (status) {
253
- case 'PENDING':
254
- return 'neutral';
255
- case 'RUNNING':
256
- return 'info';
257
- case 'FINISHED':
258
- return 'success';
259
- case 'ERROR':
260
- return 'warning';
261
- case 'SKIPPED':
262
- return 'neutral';
263
- }
264
- }
265
-
266
- function formatDuration(ms?: number): string {
267
- if (ms === undefined) return '—';
268
- if (ms < 1000) return ms + 'ms';
269
- const s = ms / 1000;
270
- if (s < 60) return s.toFixed(1) + 's';
271
- const m = Math.floor(s / 60);
272
- const rem = Math.round(s - m * 60);
273
- return m + 'm ' + rem + 's';
274
- }
275
-
276
- function elapsed(state: RunState): number {
277
- const end = state.finishedAt ?? Date.now();
278
- return end - state.startedAt;
279
- }
280
-
281
- function DAGGraph({ state }: { state: RunState }): JSX.Element {
282
- const theme = useHostTheme();
283
- const layout = computeDAGLayout({
284
- nodes: state.tasks.map((t) => ({ id: t.id })),
285
- edges: state.tasks.flatMap((t) =>
286
- t.depends_on.map((d) => ({ from: d, to: t.id })),
287
- ),
288
- direction: 'vertical',
289
- nodeWidth: NODE_W,
290
- nodeHeight: NODE_H,
291
- rankGap: 72,
292
- nodeGap: 40,
293
- padding: 24,
294
- });
295
- const byId = new Map(state.tasks.map((t) => [t.id, t]));
296
-
297
- return (
298
- <svg width={layout.width} height={layout.height} style={{ display: 'block', maxWidth: '100%' }}>
299
- {layout.edges.map((e, i) => (
300
- <line
301
- key={'e' + i}
302
- x1={e.sourceX}
303
- y1={e.sourceY}
304
- x2={e.targetX}
305
- y2={e.targetY}
306
- stroke={theme.stroke.secondary}
307
- strokeWidth={1.25}
308
- markerEnd="url(#dag-arrow)"
309
- />
310
- ))}
311
- <defs>
312
- <marker id="dag-arrow" viewBox="0 0 10 10" refX="9" refY="5" markerWidth="6" markerHeight="6" orient="auto-start-reverse">
313
- <path d="M 0 0 L 10 5 L 0 10 z" fill={theme.stroke.secondary} />
314
- </marker>
315
- </defs>
316
- {layout.nodes.map((n) => {
317
- const t = byId.get(n.id);
318
- if (!t) return null;
319
- return (
320
- <g key={n.id} transform={'translate(' + n.x + ',' + n.y + ')'}>
321
- <rect
322
- width={NODE_W}
323
- height={NODE_H}
324
- rx={8}
325
- ry={8}
326
- fill={theme.fill.secondary}
327
- stroke={theme.stroke.secondary}
328
- strokeWidth={t.status === 'RUNNING' ? 2 : 1}
329
- />
330
- <text x={12} y={22} fontSize={13} fontWeight={600} fill={theme.text.primary}>
331
- {t.id.length > 22 ? t.id.slice(0, 21) + '…' : t.id}
332
- </text>
333
- <text x={12} y={42} fontSize={10.5} fill={theme.text.secondary}>
334
- {t.executor} · {t.complexity} · {t.status}
335
- </text>
336
- <text x={12} y={58} fontSize={10.5} fill={theme.text.tertiary}>
337
- {formatDuration(t.durationMs)}
338
- </text>
339
- </g>
340
- );
341
- })}
342
- </svg>
343
- );
344
- }
345
-
346
- export default function DagRun(): JSX.Element {
347
- const counts = useMemo(
348
- () =>
349
- STATE.tasks.reduce(
350
- (acc, t) => {
351
- acc.total += 1;
352
- const key = t.status.toLowerCase() as 'pending' | 'running' | 'finished' | 'error' | 'skipped';
353
- if (key in acc) acc[key] += 1;
354
- return acc;
355
- },
356
- { total: 0, pending: 0, running: 0, finished: 0, error: 0, skipped: 0 },
357
- ),
358
- [],
359
- );
360
- const isFinal = STATE.finishedAt !== undefined;
361
- const statusLabel =
362
- STATE.runOutcome === 'PAUSED'
363
- ? 'PAUSED'
364
- : STATE.runOutcome === 'FAILED'
365
- ? 'FAILED'
366
- : isFinal
367
- ? 'COMPLETE'
368
- : 'RUNNING';
369
- const statusTone =
370
- STATE.runOutcome === 'FAILED'
371
- ? 'danger'
372
- : STATE.runOutcome === 'PAUSED'
373
- ? 'warning'
374
- : isFinal
375
- ? 'success'
376
- : 'info';
377
-
378
- return (
379
- <Stack gap={20}>
380
- <Stack gap={6}>
381
- <H1>{STATE.title}</H1>
382
- <Text tone="secondary" size="small">
383
- run {STATE.runId}
384
- </Text>
385
- <Row gap={10} align="center">
386
- <Pill tone={statusTone} active size="sm">
387
- {statusLabel}
388
- </Pill>
389
- <Text tone="secondary" size="small">
390
- {counts.total} tasks · elapsed {formatDuration(elapsed(STATE))}
391
- </Text>
392
- </Row>
393
- {STATE.runMessage ? (
394
- <Text tone="secondary" size="small">
395
- {STATE.runMessage}
396
- </Text>
397
- ) : null}
398
- </Stack>
399
-
400
- <Grid columns={6} gap={12}>
401
- <Stat value={String(counts.total)} label="Total" />
402
- <Stat value={String(counts.pending)} label="Pending" />
403
- <Stat value={String(counts.running)} label="Running" tone={counts.running > 0 ? 'info' : undefined} />
404
- <Stat value={String(counts.finished)} label="Finished" tone={counts.finished > 0 ? 'success' : undefined} />
405
- <Stat value={String(counts.error)} label="Errored" tone={counts.error > 0 ? 'danger' : undefined} />
406
- <Stat value={String(counts.skipped)} label="Skipped" />
407
- </Grid>
408
-
409
- <Divider />
410
-
411
- <Stack gap={12}>
412
- <H2>Graph</H2>
413
- <DAGGraph state={STATE} />
414
- </Stack>
415
-
416
- <Divider />
417
-
418
- <Stack gap={12}>
419
- <H2>Tasks</H2>
420
- <Stack gap={10}>
421
- {STATE.tasks.map((t) => (
422
- <Card key={t.id} collapsible defaultOpen={t.status !== 'PENDING'}>
423
- <CardHeader
424
- trailing={
425
- <Row gap={6} align="center">
426
- <Pill tone="neutral" size="sm">
427
- {t.executor}
428
- </Pill>
429
- <Pill tone={pillToneFor(t.status)} active={t.status !== 'PENDING'} size="sm">
430
- {t.status}
431
- </Pill>
432
- </Row>
433
- }
434
- >
435
- {t.id}
436
- </CardHeader>
437
- <CardBody>
438
- <Stack gap={8}>
439
- <Text tone="secondary" size="small">
440
- Model {t.model}
441
- {t.depends_on.length > 0 ? ' · depends on ' + t.depends_on.join(', ') : ''}
442
- {t.durationMs !== undefined ? ' · ' + formatDuration(t.durationMs) : ''}
443
- </Text>
444
- <Text size="small">
445
- <Text as="span" weight="semibold">
446
- Prompt:{' '}
447
- </Text>
448
- {t.subtask_prompt}
449
- </Text>
450
- {t.resultText ? (
451
- <pre style={{ margin: 0, whiteSpace: 'pre-wrap', wordBreak: 'break-word' }}>
452
- {t.resultText}
453
- {t.status === 'RUNNING' ? '\u2588' : ''}
454
- </pre>
455
- ) : null}
456
- {t.skippedReason ? (
457
- <Text size="small" tone="secondary">
458
- Skipped: {t.skippedReason}
459
- </Text>
460
- ) : null}
461
- {t.errorMessage ? (
462
- <Text size="small" tone="secondary">
463
- {t.errorMessage}
464
- </Text>
465
- ) : null}
466
- </Stack>
467
- </CardBody>
468
- </Card>
469
- ))}
470
- </Stack>
471
- </Stack>
472
- </Stack>
473
- );
248
+ const CANVAS_BODY = String.raw `const NODE_W = 200;
249
+ const NODE_H = 72;
250
+
251
+ function pillToneFor(status: TaskStatus): 'neutral' | 'info' | 'success' | 'warning' {
252
+ switch (status) {
253
+ case 'PENDING':
254
+ return 'neutral';
255
+ case 'RUNNING':
256
+ return 'info';
257
+ case 'FINISHED':
258
+ return 'success';
259
+ case 'ERROR':
260
+ return 'warning';
261
+ case 'SKIPPED':
262
+ return 'neutral';
263
+ }
264
+ }
265
+
266
+ function formatDuration(ms?: number): string {
267
+ if (ms === undefined) return '—';
268
+ if (ms < 1000) return ms + 'ms';
269
+ const s = ms / 1000;
270
+ if (s < 60) return s.toFixed(1) + 's';
271
+ const m = Math.floor(s / 60);
272
+ const rem = Math.round(s - m * 60);
273
+ return m + 'm ' + rem + 's';
274
+ }
275
+
276
+ function elapsed(state: RunState): number {
277
+ const end = state.finishedAt ?? Date.now();
278
+ return end - state.startedAt;
279
+ }
280
+
281
+ function DAGGraph({ state }: { state: RunState }): JSX.Element {
282
+ const theme = useHostTheme();
283
+ const layout = computeDAGLayout({
284
+ nodes: state.tasks.map((t) => ({ id: t.id })),
285
+ edges: state.tasks.flatMap((t) =>
286
+ t.depends_on.map((d) => ({ from: d, to: t.id })),
287
+ ),
288
+ direction: 'vertical',
289
+ nodeWidth: NODE_W,
290
+ nodeHeight: NODE_H,
291
+ rankGap: 72,
292
+ nodeGap: 40,
293
+ padding: 24,
294
+ });
295
+ const byId = new Map(state.tasks.map((t) => [t.id, t]));
296
+
297
+ return (
298
+ <svg width={layout.width} height={layout.height} style={{ display: 'block', maxWidth: '100%' }}>
299
+ {layout.edges.map((e, i) => (
300
+ <line
301
+ key={'e' + i}
302
+ x1={e.sourceX}
303
+ y1={e.sourceY}
304
+ x2={e.targetX}
305
+ y2={e.targetY}
306
+ stroke={theme.stroke.secondary}
307
+ strokeWidth={1.25}
308
+ markerEnd="url(#dag-arrow)"
309
+ />
310
+ ))}
311
+ <defs>
312
+ <marker id="dag-arrow" viewBox="0 0 10 10" refX="9" refY="5" markerWidth="6" markerHeight="6" orient="auto-start-reverse">
313
+ <path d="M 0 0 L 10 5 L 0 10 z" fill={theme.stroke.secondary} />
314
+ </marker>
315
+ </defs>
316
+ {layout.nodes.map((n) => {
317
+ const t = byId.get(n.id);
318
+ if (!t) return null;
319
+ return (
320
+ <g key={n.id} transform={'translate(' + n.x + ',' + n.y + ')'}>
321
+ <rect
322
+ width={NODE_W}
323
+ height={NODE_H}
324
+ rx={8}
325
+ ry={8}
326
+ fill={theme.fill.secondary}
327
+ stroke={theme.stroke.secondary}
328
+ strokeWidth={t.status === 'RUNNING' ? 2 : 1}
329
+ />
330
+ <text x={12} y={22} fontSize={13} fontWeight={600} fill={theme.text.primary}>
331
+ {t.id.length > 22 ? t.id.slice(0, 21) + '…' : t.id}
332
+ </text>
333
+ <text x={12} y={42} fontSize={10.5} fill={theme.text.secondary}>
334
+ {t.executor} · {t.complexity} · {t.status}
335
+ </text>
336
+ <text x={12} y={58} fontSize={10.5} fill={theme.text.tertiary}>
337
+ {formatDuration(t.durationMs)}
338
+ </text>
339
+ </g>
340
+ );
341
+ })}
342
+ </svg>
343
+ );
344
+ }
345
+
346
+ export default function DagRun(): JSX.Element {
347
+ const counts = useMemo(
348
+ () =>
349
+ STATE.tasks.reduce(
350
+ (acc, t) => {
351
+ acc.total += 1;
352
+ const key = t.status.toLowerCase() as 'pending' | 'running' | 'finished' | 'error' | 'skipped';
353
+ if (key in acc) acc[key] += 1;
354
+ return acc;
355
+ },
356
+ { total: 0, pending: 0, running: 0, finished: 0, error: 0, skipped: 0 },
357
+ ),
358
+ [],
359
+ );
360
+ const isFinal = STATE.finishedAt !== undefined;
361
+ const statusLabel =
362
+ STATE.runOutcome === 'PAUSED'
363
+ ? 'PAUSED'
364
+ : STATE.runOutcome === 'FAILED'
365
+ ? 'FAILED'
366
+ : isFinal
367
+ ? 'COMPLETE'
368
+ : 'RUNNING';
369
+ const statusTone =
370
+ STATE.runOutcome === 'FAILED'
371
+ ? 'danger'
372
+ : STATE.runOutcome === 'PAUSED'
373
+ ? 'warning'
374
+ : isFinal
375
+ ? 'success'
376
+ : 'info';
377
+
378
+ return (
379
+ <Stack gap={20}>
380
+ <Stack gap={6}>
381
+ <H1>{STATE.title}</H1>
382
+ <Text tone="secondary" size="small">
383
+ run {STATE.runId}
384
+ </Text>
385
+ <Row gap={10} align="center">
386
+ <Pill tone={statusTone} active size="sm">
387
+ {statusLabel}
388
+ </Pill>
389
+ <Text tone="secondary" size="small">
390
+ {counts.total} tasks · elapsed {formatDuration(elapsed(STATE))}
391
+ </Text>
392
+ </Row>
393
+ {STATE.runMessage ? (
394
+ <Text tone="secondary" size="small">
395
+ {STATE.runMessage}
396
+ </Text>
397
+ ) : null}
398
+ </Stack>
399
+
400
+ <Grid columns={6} gap={12}>
401
+ <Stat value={String(counts.total)} label="Total" />
402
+ <Stat value={String(counts.pending)} label="Pending" />
403
+ <Stat value={String(counts.running)} label="Running" tone={counts.running > 0 ? 'info' : undefined} />
404
+ <Stat value={String(counts.finished)} label="Finished" tone={counts.finished > 0 ? 'success' : undefined} />
405
+ <Stat value={String(counts.error)} label="Errored" tone={counts.error > 0 ? 'danger' : undefined} />
406
+ <Stat value={String(counts.skipped)} label="Skipped" />
407
+ </Grid>
408
+
409
+ <Divider />
410
+
411
+ <Stack gap={12}>
412
+ <H2>Graph</H2>
413
+ <DAGGraph state={STATE} />
414
+ </Stack>
415
+
416
+ <Divider />
417
+
418
+ <Stack gap={12}>
419
+ <H2>Tasks</H2>
420
+ <Stack gap={10}>
421
+ {STATE.tasks.map((t) => (
422
+ <Card key={t.id} collapsible defaultOpen={t.status !== 'PENDING'}>
423
+ <CardHeader
424
+ trailing={
425
+ <Row gap={6} align="center">
426
+ <Pill tone="neutral" size="sm">
427
+ {t.executor}
428
+ </Pill>
429
+ <Pill tone={pillToneFor(t.status)} active={t.status !== 'PENDING'} size="sm">
430
+ {t.status}
431
+ </Pill>
432
+ </Row>
433
+ }
434
+ >
435
+ {t.id}
436
+ </CardHeader>
437
+ <CardBody>
438
+ <Stack gap={8}>
439
+ <Text tone="secondary" size="small">
440
+ Model {t.model}
441
+ {t.depends_on.length > 0 ? ' · depends on ' + t.depends_on.join(', ') : ''}
442
+ {t.durationMs !== undefined ? ' · ' + formatDuration(t.durationMs) : ''}
443
+ </Text>
444
+ <Text size="small">
445
+ <Text as="span" weight="semibold">
446
+ Prompt:{' '}
447
+ </Text>
448
+ {t.subtask_prompt}
449
+ </Text>
450
+ {t.resultText ? (
451
+ <pre style={{ margin: 0, whiteSpace: 'pre-wrap', wordBreak: 'break-word' }}>
452
+ {t.resultText}
453
+ {t.status === 'RUNNING' ? '\u2588' : ''}
454
+ </pre>
455
+ ) : null}
456
+ {t.skippedReason ? (
457
+ <Text size="small" tone="secondary">
458
+ Skipped: {t.skippedReason}
459
+ </Text>
460
+ ) : null}
461
+ {t.errorMessage ? (
462
+ <Text size="small" tone="secondary">
463
+ {t.errorMessage}
464
+ </Text>
465
+ ) : null}
466
+ </Stack>
467
+ </CardBody>
468
+ </Card>
469
+ ))}
470
+ </Stack>
471
+ </Stack>
472
+ </Stack>
473
+ );
474
474
  }`;
@@ -0,0 +1,14 @@
1
+ import { structuredContractOutputSchemaIds, } from "./types.js";
2
+ const validators = new Map();
3
+ export function registerStructuredContractValidator(schemaId, validator) {
4
+ if (validators.has(schemaId)) {
5
+ throw new Error(`structured contract validator already registered for ${schemaId}`);
6
+ }
7
+ validators.set(schemaId, validator);
8
+ }
9
+ export function getStructuredContractValidator(schemaId) {
10
+ if (structuredContractOutputSchemaIds.includes(schemaId)) {
11
+ return validators.get(schemaId);
12
+ }
13
+ return undefined;
14
+ }
@@ -0,0 +1,8 @@
1
+ /**
2
+ * Side-effect module: registers every structured contract validator against
3
+ * its schemaId in the contract output registry. Imported once by the node
4
+ * executor; keeps per-contract imports out of the executor.
5
+ */
6
+ import { validateFrontendContractNodeOutput } from "./frontend-implementation-contract.js";
7
+ import { registerStructuredContractValidator } from "./contract-output-registry.js";
8
+ registerStructuredContractValidator("frontend-implementation-contract-v1", validateFrontendContractNodeOutput);
@@ -31,7 +31,15 @@ export function getNodeOutputAsJson(state, nodeId) {
31
31
  export function normalizeVerdictCandidateLine(value) {
32
32
  const trimmed = String(value).trim();
33
33
  const emphasized = trimmed.match(/^\*{1,3}\s*(VERDICT:[^*]+?)\s*\*{1,3}$/);
34
- return (emphasized ? emphasized[1] : trimmed).trim();
34
+ const line = (emphasized ? emphasized[1] : trimmed).trim();
35
+ // Canonicalize the colon separator and trailing sentence punctuation so
36
+ // "VERDICT:pass", "VERDICT: pass", or "VERDICT: pass." match the declared
37
+ // enum lines without accepting free-text verdicts. Conflicting verdicts
38
+ // still canonicalize to distinct enum lines and remain fail-closed.
39
+ const verdictMatch = line.match(/^VERDICT:\s*(pass|request-revision)\s*[.!?。!?]*$/i);
40
+ if (verdictMatch)
41
+ return `VERDICT: ${verdictMatch[1].toLowerCase()}`;
42
+ return line;
35
43
  }
36
44
  /** First line matching /^VERDICT:/ from node assistantText/stdout (markdown emphasis OK). */
37
45
  export function extractFirstVerdictLineFromNodeText(text) {